Publication

Abstract : The recommendation of products of interest to the user is pivotal for improving a customer's shopping experience. Recommender system has diversified and endeared itself in wide ranging industrial applications from e-commerce to online video sites. As the input data that is supplied to the recommender systems is large, the recommender system is often considered as data intensive application. In this paper, we present improvised MapReduce based data preprocessing and content based recommendation algorithms. Also, Spark based content based recommendation algorithm is developed and compared with Hadoop based content based recommendation algorithm. Our experimental results on Amazon co-purchasing network meta data show that Spark based content based recommendation algorithm is faster than Hadoop based content based recommendation algorithm. Also, graphical user interface is developed to interact with the recommender system.